An experimental evaluation of the cache partitioning impact on multicore real-time schedulers

Abstract

Shared cache partitioning is a well-known technique used in multicore real-time systems to isolate task workloads and improve system predictability. Presently, the state-of-the-art studies that evaluate shared cache partitioning on multicore processors lack two key issues. First, the cache partitioning mechanism is typically implemented either in a simulation environment or in a general-purpose OS, and so the impact of kernel activities, such as interrupt handlers and context switching, on the task partitions tend to be overlooked. Second, the evaluation is typically restricted to either a global or partitioned scheduler, thereby by falling to compare the performance of cache partitioning when tasks are scheduled by different schedulers. In this work, we design and implement a shared cache partitioning mechanism in a multicore component-based RTOS capable of assigning partitions to internal OS data structures, including task and system stacks and interrupt handlers data. We evaluate our shared cache partitioning mechanism running task sets under global (G-EDF) and partitioned (P-EDF) multicore real-time scheduling algorithms. Our results indicate that a lightweight RTOS does not impact real-time tasks, and shared cache partitioning has different behavior depending on the scheduler and the task's working set size.